Indicators: advanced averages and trend (DEMA, KAMA, Supertrend)

What it does

This second wave completes the base averages without overlapping them. There is no Source select here: every study is defined on the close (or on high/low/close for SuperTrend). DEMA is 2·EMA − EMA(EMA); TEMA is 3·e1 − 3·e2 + e3; both use the catalog EMA (SMA seed, alpha = 2/(N+1)) and each extra EMA stage runs on the defined suffix of the previous one, so DEMA is defined from index 2(N−1) and TEMA from 3(N−1).

ZLEMA de-lags the close with adj = 2·close − close[i − lag], lag = floor((N−1)/2), then applies an EMA. VWMA is Σ(close·volume)/Σ(volume) over N bars and returns no value when the window volume is zero. KAMA follows Kaufman: efficiency ratio over N, smoothing constant squared between the fast and slow EMA constants, seeded with the raw close. ALMA applies a Gaussian window shifted by offset with width N/σ. McGinley Dynamic uses the TradingView/ATAS variant without the 0.6 factor, seeded with the SMA. LSMA is the end point of a least-squares fit over N bars.

Linear Regression Channel is the rolling version: at each bar the fit is redone and the terminal point ± k·σ of residuals (population σ) is emitted as a band. SuperTrend uses Wilder ATR on hl2 with the usual ratchet; it draws two lines, the lower one in an uptrend and the upper one in a downtrend, each empty outside its trend. Price Envelopes draw SMA(N) × (1 ± p/100).

Settings

How to use it

Enable a study from the Library groups of the Indicators panel; the same search field filters by label and group. Every study writes on the price scale, so several can be stacked on the same chart. SuperTrend flips are visible as one line stopping and the other starting; no vertical connector is drawn at the flip, because continuity only exists inside a trend.

Trend start for SuperTrend is chosen at the first defined ATR: uptrend if close ≥ hl2, downtrend otherwise. This differs from platforms that always start bullish, which avoids a first bar drawn against the price.

Limits and pitfalls

• Warm-up compounds: TEMA(20) has no value before bar 57, and ZLEMA(20) before lag + N − 1 = 28.
• VWMA returns no point when the window volume is zero (placeholder bars without trades); 0/0 is treated as missing, not as zero.
• McGinley skips its update when the adaptive denominator is not finite or not positive, which needs a zero or non-positive price or basis; the previous value is then kept.
• The regression channel is a rolling series, not a static channel redrawn over the window. It traces the right edge of the channel through time.
• None of these studies read the session anchor or the tick size; an unknown tick size does not change their output.